Types Of Affect Mental Status Exam
Search Results:
- When a thought disorder exists, associations between thoughts may be disconnected, constantly changing, or blocked. The person may use vague, nonspecific terms that indicate an impoverishment of content or he or she may have difficulty with abstract...
- This is important information for determining whether the person has organic mental impairment. Do you know the day of the week? What month is it? Do you know where you are? Do you know who I am? Do you remember your name? These questions are...
- Social judgement is also evaluated. A question commonly used is "If you were to find a stamp, addressed envelope lying on the sidewalk, what would you do? She is slightly disheveled. She is cooperative with the interviewer and is judged to be an adequate historian. Her mood and affect are depressed and anxious. She became tearful throughout the interview. Her flow of thought is coherent and her thought content reveals feelings of low self-esteem as well as auditory hallucinations that are self-demeaning. She admits to suicidal ideas but denies active plan or intent.
- Her orientation is good. She knows the current date, place, and person. Recent and remote memory are good. Fund of knowledge is adequate. The client shows some insight and judgment regarding her illness and need for help. Excerpted from Mabbett, P. Albany, NY: Delmar Publishers.
- However, this common practice is often widely inconsistent between raters. Recent advances in the field of Facial Action Recognition FAR have enabled the development of tools that can act to identify facial expressions from videos. In this study, we aimed to explore the potential of using machine learning techniques on FAR features extracted from videotaped semi-structured psychiatric interviews of 25 male schizophrenia inpatients mean age Then, a novel computer vision algorithm and a machine learning method were used to predict affect classification based on each psychiatrist affect rating. The algorithm is shown to have a significant predictive power for each of the human raters. This study serves as a proof-of-concept for the potential of using the machine learning FAR system as a clinician-supporting tool, in an attempt to improve the consistency and reliability of mental status examination.
- Introduction The determination of mood and affect is a major aspect of the psychiatric mental status examination MSE 1 , 2. However, in contrast to other aspects of the mental status, mood and affect determination is often challenging for clinicians, especially during the initial part of their careers 3 — 6. Few studies to date have described the inconsistent apprehension of what is mood and affect between clinicians 5 — 7. The inconsistency in the way different clinicians understand the concept of mood and affect consequently contributes to differences in the identification of affect between clinicians 8. This finding underscores the need to develop reliable tools that may serve as an aid to clinicians when conducting a psychiatric mental status examination, a concept that was presented long ago 9 , 10 , but is gaining interest as science advances and as machine learning discipline shows the capacity of developing decision supporting tools for psychiatrists Computer face recognition systems allow identification of a person from a digital image or a video frame.
Ten Point Guide To Mental State Examination (MSE) In Psychiatry
Though it has been mostly used in the context of security and surveillance, the applications of such technology are far more diverse In medicine, applications of facial recognition include the identification of rare genetic disorders that result in unique facial features 13 , An even newer field is Face Action Recognition FAR , a research domain that has made great strides over the past decade In some real-world applications, the goal is to recognize or infer intention or other psychological states rather than facial actions alone. There are many applications in the FAR field, including human-computer interfaces, video surveillance and patient condition monitoring. A FAR system is normally composed of four main steps: i face detection and tracking, ii face alignment-which is based on locating semantic facial landmarks such as the eyes, nose, mouth and chin iii feature extraction, and iv classification.- While steps i — iii are done with computer vision techniques, step iv is done with machine learning algorithms such as the Support Vectors Machine SVM algorithm. In the current study, we applied FAR machine learning algorithms to classify affects based on the classification of five psychiatrists that rated the affects of 25 male schizophrenia patients that underwent a videotaped semi-structured psychiatric interview. We aimed to achieve two goals: 1 to show that in a clinically homogenous patient population, the rating of affect is highly heterogenous among experienced clinicians and 2 to employ a unique, novel FAR system that we had previously developed 16 in an attempt to identify affect in the MSE, which may in turn lead to harmonization and better standardization of the MSE.
- The investigation was carried out in accordance with the latest version of the Declaration of Helsinki and the study design was reviewed and approved by the institutional review board. Written informed consent of the participants was obtained after the nature of the procedures had been fully explained. Demographic data of study subjects was obtained from the medical records and can be found in Supplemental Data. All interviews lasted between 7 and 10 minutes. All patients were interviewed by the lead author RB , a resident psychiatrist, in a semi-structured interview composed of the following 10 questions: 1 Can you please present yourself and tell me a bit about yourself? Measures of Affect Annotation of videos was performed for three domains of affect, as described in the psychiatric MSE chapter of a classic textbook of Psychiatry 2. Quality of affect was annotated as either dysphoric, euthymic or manic. Range of affect was classified as full, restricted, flat or blunt. Subtype of affect was categorized as one of the twelve: stupid, euphoric, empathetic, self-contemptuous, anxious, suspicious, hopeless, frightened, irritable, with a sense of guilt or other 2.
- Face Recognition Procedures The quality of the extracted features representing the video content plays a key role for the classification task that is required for the affect classification. A detailed description of the image processing and computational complexity can be found in our previous report In brief, the proposed system is designed to be reliable and rapid and support real-time analysis applications.
- The method is based on capturing local changes and encoding these local motions into a histogram of frequencies. In this approach, a face video is modeled as a sequence of histograms by the following procedure: 1 An input face image aligned and is cropped to generate a normalized face image Figure 1 ; 2 Each normalized frame is divided into a grid of equally sized cells; 3 The mean gray level intensity of each cell r at frame n is recorded as Xr[n].
- Top The input frames 1 , 4 , 5 , 17 — 19 ; Middle Normalized face; Bottom The mean gray level intensity of left eye cell along time domain. Image was taken from painDB dataset 20 , a publically available data source. The person in the figure is not related to the study. For the analysis of which face part contributes the most for the affect annotation, we divided the facial grid to six parts: left and right eye, left and right cheek, mouth and nose Figure 2. Right The face divided to six parts: left and right eye, left and right cheek, mouth and nose. Image was taken from FEED database 22 , a publically available data source. We therefore devised three mid-level features, which capture the facial expression scores, the dominating expression labels, and the facial motion.
- These vectors have previously been described Briefly: 1 Expression - for each frame in a video, we predicted the score of seven expressions norm, anger, disgust, fear, happiness, sadness and surprise , by employing a classifier that is based on the recorded mean intensities, denoted by Xr[n] above. In order to describe a video, we aggregated the expression scores to one vector of length 7 that describes a video. This calculation simply captures the variance of the face in a video. In addition, since in some cases a label might be more evident in one vector type than in another, we also consider the combination of the different vectors. These are obtained by concatenating 4 Motion and label, or 5 Motion and expression. Classifiers A linear support vector machine SVM was used as the classifier for the proposed method.
- The basic SVM method outputs binary decisions, and the multiclass classification required for our application three classes for quality; four classes for range; twelve classes for subtype is accomplished by using the one-versus-all rule. During training, a classifier is learned to separate each class from the rest. For example, for the quality classifier, a first binary classifier treats dysphoric cases as positive cases, and euthymic or manic cases as negative cases, a second classifier labels only euthymic cases as positive, the rest as negative, and a third classifier labels only manic cases as positive.
The Mental Status Examination - Clinical Methods - NCBI Bookshelf
Given a sample to classify, each classifier outputs a score that is aimed to be negative if the sample is from the negative group, and positive for a sample that is associated with the positive label. Once training is complete, a test video is assigned to the label of the binary classifier with the highest response. In our experiments, the classifiers are employed as part of a Leave One Out procedure, as explained in the next section. Statistical Methods We perform statistical analysis in order to assess the following: i the agreement between the five raters, ii the predictive power of the automatic method, iii the predictive power per facial part, and iv the success of the machine learning procedure in predicting the rating of each rater. For measuring the agreement between a pair of raters, we consider the ratings given by the two over all cases. We then record the ratio of the cases in which the two provided the same annotation. In order to measure the predictive power of each one of the three types of vectors and two combinations thereof , we employ a logistic regression analysis.- This is done separately for each rater and to each vector type the five types described in Sec. Each separate analysis a total of 5 types of vectors, including the 2 hybrid types, times 5 raters, times 3 domains, or a total of 75 experiments employs the classes of the domain as dependent variables. In each experiment, we perform multinomial logistic regression using the R version 0. The p-value reflects the Likelihood ratio test vs. In order to compare the predictive power of each facial part separately, these experiments are then repeated for the Motion descriptor only the third feature vector in Sec. The other features require the classification of the facial expression, which in turn is based on the entire face. Here, too, the p-values are obtained using multinomial logistic regression. While the multinomial regression experiments provide statistical significance, we employ Support Vector Machines SVM in order to estimate the classification accuracy, i. This is done to each rater separately, due to the lack of consensus.
Mental Status Exam (MSE) — Factors And Definitions | Medical Library
The samples of all other patients are taken as the training samples and the learned classifier is used to predict the target label of the held-out sample. This is repeated to all patients. Results Agreement between Raters We first evaluated the agreement between the five psychiatrist raters. It is immediately evident that the distributions of each of the labels vary considerably between the raters, in all three domains of affect investigated in this study: quality Table 1A , range Table 1C and subtype of affect Table 1E. The agreement between the raters was defined as the number of identical annotations between two raters divided by the number of samples. Values represents percentage of agreements of specific affect subtype classification. Logistic regression analyses P-values for each classifier for the affect quality, range and subtype are detailed in Table 2A—C. As can be seen, the quality dimension is captured in a statistically significant way, for most raters, using the motion feature vector.- The expression feature is significantly associated with subtype in at least four out of five raters. Psychiatrist 1 is missing due to lack of variance in their quality labeling. Identification of Face Parts That Are Used by Raters to Enable Affect Annotation We then proceeded to investigate which parts of the face are being used by the psychiatrist to determine the affect annotation for quality, range and sub-type. To that end, we divided the face of the subjects to six parts: left and right eye, left and right cheek, mouth and nose Figure 2 , and analyzed the classifier based on the motion features for each rater to determine the relative contribution of each face part to the annotation in each rater.
Mental Status Examination (MSE) Teaching Activities And Resources | RNAO
Tables 3 summarize the results of the Logistic regressions for face parts contribution to the affect quality Table 3A , range Table 3B and subtype Table 3C. Values represent P-values calculated for each face part. Using the features described previously and the SVM classifier, we have predicted the annotation of raters for each patient. This was done in a LOO Leave-one-out manner, in order not to make predictions on the training set. Since the distribution of the labels is not uniform, the chance level ranges between the various raters, and we therefore also report the improvement over baseline calculated as the ratio between the obtained accuracy and the frequency of the most common label for each rater 1. Shown are both the obtained accuracies and the improvement coefficient computed as the ratio between the obtained accuracy and the accuracy obtained by guessing the most common label.
Comments
Post a Comment